The Agent Skills Directory

[SAFE]: The skill demonstrates a high level of security awareness and implements multiple defensive layers to protect the user's repository and credentials.
[COMMAND_EXECUTION]: The scripts/check_env.py script uses subprocess.run to execute Git commands such as rev-parse. It safely uses list-style arguments (shell=False) and restricts its operations to read-only inspections, explicitly avoiding dependency installation or secret reading.
[INDIRECT_PROMPT_INJECTION]: The skill processes repository data including manifests, README files, and Git history. It mitigates injection risks through specific instructions for the agent to treat setup as an evidence-driven task with defined boundary markers. The scripts/init_session_start_hook.py script employs a robust blacklist of shell-sensitive characters (FORBIDDEN_GUIDE_CHARS) to sanitize user-provided guide paths before they are incorporated into shell commands within lifecycle hooks.
[DATA_EXFILTRATION]: The skill includes explicit prohibitions against reading or echoing secret values, tokens, or .env file contents. The scripts/sanitize_history.sh script is a defensive utility designed to scan Git history for sensitive patterns and guide the user through revocation and cleanup, rather than exfiltrating data.
[PERSISTENCE]: The skill facilitates the installation of SessionStart hooks for startup automation, a platform-supported feature. The implementation is guarded by mandatory --dry-run previews and requires explicit user approval before mutating configuration files.

auto-repo-setup